Context-Aware Dynamic Power Management for Self-Powered Body Sensor Networks

Body sensor networks (BSNs) have shown significant potential in health applications by empowering researchers, scientists, doctors, caregivers, and patients to explore correlations between human-related sensor data and human health through continuous, vigilant, remote, and non-invasive data collection. To enable continuous vigilant monitoring for long-term logging of sensor data without human intervention, the operation time of BSNs is a significant concern. Harvesting energy from the body and the ambient environment has become a promising solution for realizing self-powered sensor systems capable of quasi-perpetual operation. However, the discontinuous and dynamic characteristics of energy harvesting in real-world scenarios – and their implications for the design and operation of self-powered sensor systems – are not yet well studied.
Conventional characterization of energy harvesters is done in a laboratory environment, without much consideration for such real-world dynamics. In order to better understand the nature of the energy sources like solar and thermoelectric and how the human behavior affects the energy harvesting, we designed a custom Energy Harvesting and Data Collection (EHDC) platform to explore energy harvesting dynamics by longtime profiling in the real world.
Since the energy sources are uncontrollable but often predictable, we proposed a context-aware hybrid model for the multimodal indoor and outdoor energy harvesting prediction. Here we term context to refer to energy harvesting related factors including environmental parameters like light intensity, temperature, weather forecast; and human behavior like motion, schedule, and location. By leveraging the knowledge of the current context and near future, the system could predict harvested energy more accurately and thus improve the efficiency of power consumption.
The core of this work is the design of the context-aware dynamic power management framework for self-powered body sensor networks. With the understanding of energy harvesting dynamics, the framework is proposed to efficiently use the harvested power to optimize the data quality according to the environmental and behavioral context. The power management in an energy harvesting system is formalized as a convex optimization problem, and the optimal solution is derived. Online scheduling in the real world with energy prediction is discussed. A case study of Atrial Fibrillation detection is analyzed to demonstrate the application-specific utility/cost function which could better represent the demand of the application. Such application-specific cost analysis outperforms the methodologies that solely from the perspective of digital signal processing, or generally assume a linear cost function in related work which would be too simplified.
We validate this framework by designing a custom Self-powered and Context-aware Dynamic Power Management (SCDPM) platform. The platform is capable of vigilant health monitoring with ECG and motion data. It also collects environmental data which help to understand the context and make dynamic power management. The SCDPM is an ultralow power platform which performs better than state-of-the-art health monitoring platforms regarding system power consumption and the dynamic power management and adaptive sensing capabilities.
Overall, this work explores the energy harvesting dynamics to improve the design of self-power sensor systems and the proposed context-aware dynamic power management framework improves the self-powered BSN performance and operation time by taking advantage of context information.